Research On Target Trajectory Semantic Analysis And Location Prediction Based On Chinese Semantic Slot Filling

With the rapid development of communication technology and positioning technology,moving target trajectory semantic analysis and location prediction have more and more important significance on research,which is in daily traveling,warning of national defense and epidemic prevention and control.At present,most of the existing studies carry out data mining according to the motion law of the motion trajectory itself.However,the target motion trajectory contains not only the information of target state,but also the information of interaction behavior between the target and the surrounding environment.As we know,language is the vital information carrier when people communicate and deal with natural environment message.Therefore,this thesis studies the semantic analysis and location prediction of target trajectory from the semantic level.The main work of this thesis includes:(1)An enhancement method for Chinese environment semantic slot filling based on POS tagging.Considering the different requirements,as well as the complexity of geographical nouns,for semantic feature extraction in different trajectory semantic analysis and location prediction tasks,this thesis uses word segmentation method based on dictionary and statistics to process corpus information.Then POS tags are used to fuzzify the model input sequence,and POS tags and dependency parsing are used to optimize the sequence annotation algorithm.According to the experimental results,this method not only improves the model training efficiency and slot filling accuracy,but also enhances the universal applicability of the model in different scenarios,which lays a foundation for the language information processing of trajectory semantic analysis and position prediction tasks.(2)Trajectory language description generation method based on target behavior analysis.Firstly,the behavior semantic modeling and expression of moving targets are obtained by modeling the target trajectory unit and calculating the topological behavior.Secondly,the basic words are extracted by Chinese word segmentation and classified by semantic slot filling method,combined with the hierarchical clustering method.Then the description language framework is extracted according to the topological behavior,and finally the trajectory data is transformed from digital information to language information.(3)Location prediction method based on slot filling and trajectory semantics.The trajectory grid method is used to simplify the target trajectory.At the same time,the semantic slot filling algorithm is used to extract the possible activity types of the target at each location point.The topological semantics are obtained through topological behavior analysis.The two semantic information are combined with dwell point and dwell time for multimodal embedding,and the semantic multimodal embedding location prediction model is obtained.Finally,it gives experiments to show the effectiveness of this method and the accuracy of location prediction.